Statistical process control of path-planning algorithms for robots based on visual feedback

Abstract

This work presents an analysis of the effects of implementing curves generated by B-Spline and Catmull-Rom splines on a path-planning algorithm for a 6-axis robot. The objective of this project is to validate the use of path-planning algorithms based on curve fitting by splines to maintain a Process Capability Ratio CpK of at least 1.33. The project serves as a first approach to building a working alternative for an edge-grinding system for glass workpieces produced in the automotive industry. The current manufacturing process is done by a CNC machine, but it lacks the flexibility needed to keep up with the changing market needs generated by Industry 4.0. The setup built consisted of a Cognex In-Sight camera as a Vision System to detect and recognize the shape of the workpieces used, MATLAB to filter, process, and apply the curve fitting algorithms to the detected points to generate a tool path, and a UR5 robot to handle the workpiece, using a dial indicator tool as a remote TCP to measure the offset between the expected and the actual position of the robot. A Design of Experiments was done to determine what was the best combination of parameters that approximated the measurements better to the expected value and reduced the variance in the data. The results showed in the vast majority that the splines were capable of maintaining the position of the robot within the ± 0.4 mm tolerance requirements, however, a tolerance of ± 1 mm was chosen to assure the expected Process Capability Ratio CpK, as it is still a good tolerance range accepted for these types of applications. The variance of the measurements was different depending on the combination of parameters used, however, the best combination found was a Catmull-Rom spline and distance between points of 2 mm, and the process was capable of maintaining a CpK of 1.34. These results show that the program and installation developed are able to keep good positioning for the robot in the different paths generated.